[c-nsp] L3VPN - redistribute into OSPF, default metric

[Adam Vitkovsky]

Hi Ulrik,

Yes please, you can use "set med igp-cost" in route-policy on eBGP sessions
towards CE2 and CE3. 
This will automatically set MED value equal to the IGP metric towards the
BGP next-hop (PE1's loopback) of the CE1's prefix. 

adam
-----Original Message-----
From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of
Ulrik Ivers
Sent: Tuesday, December 10, 2013 4:11 PM
To: cisco-nsp at puck.nether.net
Subject: [c-nsp] L3VPN - redistribute into OSPF, default metric

Hi,

I have a question regarding redistributing routes from BGP to OSPF in a
L3VPN setup. I've got limited experience in this area, so please excuse if
the question is incorrectly formulated.

The topology is like this
   P -- PE1 -- CE1
   |     |
 PE2 - PE3
   |\   /|
   | \ / |
   |  X  |
   | / \ |
CE2 - CE3

A subnet is directly connected to PE1 from CE1, no OSPF. PE1 is doing
    address-family ipv4 vrf CUST-A
         redistribute connected
    exit-address-family

PE2 and PE3 (ASR9K) are doing OSPF with CE2 and CE3. PE2 and PE3 are
redistributing from BGP into OSPF, today no route-maps etc.

Now to the question. The subnet from CE1 is redistributed to CE2 and CE3
from both PE2 and PE3 with the same metric (metric 1, default for BGP into
OSPF) . The result is that CE2 and CE3 have to equal cost routes to the
subnet to both PE2 and PE3. Wouldn't it better if the route to PE2 had a
higher metric so that when all links are up the traffic goes via only PE3?

Can we accomplish this in some "automatic" way (we have several VRFs set up
in the same way)? Getting PE2 and PE3 to take into account the IGP cost to
reach PE1 when redistributing? Or do we need to manually create route-maps
in PE2 to set a higher cost for the subnet from CE1 when redistributing into
OSPF?

Regards,
/Ulrik Ivers



_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/